By Daniel Lanyon on Monday 5 November 2018
We caught up with Le Luel at the AltFi Amsterdam Summit 2018 to find out how the firm is using data science to help its lending to small businesses looking to grow.
Q: What are the cutting edge technologies transforming credit underwriting now and what are emerging?
A: I see three main emerging trends in credit underwriting:
1. The increasing use of machine learning algorithms to improve the power of risk models
2. The integration of transaction-level bank data, through bank account APIs, within credit decisions
3. The use of digital behaviour data to better assess the risk of borrowers with limited credit history
Q: To what extent is 'AI' & machine learning revolutionising lending in the SME space?
A: Machine learning models differ from traditional regression models by being significantly better at capturing diversity in risk patterns. This becomes highly relevant when assessing wide-ranging and diverse asset classes, like small business loans.
However, machine learning models also come with some weaknesses, for example in the transparency of decisions and unintended biases. This is why machine learning is being implemented gradually in the SME lending space - as lenders find ways to extract its benefits while controlling for its drawbacks.
Q: How far can automation go - will you always need a significant element of human attention?
A: Human expertise and interpretation will always be needed when designing decision models; for example to make sense of model assumptions, implement additional judgment controls and perform quality controls. This also applies when making the credit decisions themselves, although the level of human intervention needed depends on the quality of data available to train the models. For example, ventures in new markets will typically need significant levels of human intervention until enough experience has been accumulated to create strong models.
Q: How does Funding Circle's methodology in managing risk translate into its different geographies?
A: At Funding Circle we have used data science to develop proprietary risk models and methodologies when assessing small businesses. We deploy these consistently across all geographies by our global risk team. Within these models, specific algorithms will vary depending on local data, but the general decision architecture remains consistent. We also have identical risk governance practices and controls globally. This consistent approach to risk management gives us competitive advantages in terms of know-how, economies of scale, the effectiveness of controls and establishing investor trust.
Q: Has the Funding Circle/platform lending model helped you to overturn previously held ideas about creditworthiness from your former positions?
A: At Funding Circle I have experienced the power of being very focused. By choosing to only do unsecured business loans, we can analyse small business needs and behaviours in incredible depth. This provides us with a refined understanding of their creditworthiness.
As a platform, actual repayment cash flows to investors are very important to us - and these give us a clearer understanding of true returns. This is especially important when compared to banks; who adjust for complex accounting and provisioning rules when looking at portfolio performance. Overall, this experience has sharpened my vision of the dynamics of credit risk.